Fix dist transpiler develop

* refine inference_lib_dist after code move, and add it to docker/build.sh

* remove is_directory in inference_lib.cmake

Correctly handle cuda place for operators

Add Python interface of prior_boxes

* Add softmax_with_cross_entropy and smooth_l1 in Python API.

* Fix doc format.

* append cmd in manylinux dockerfile.x86

* add new line

detection map evaluator for SSD

override comparison operators in Python for Variable

Fix empty Vector foreach

Merge build docs with build using docker

rename switch_kernel.md to kernel_selection.md

* dynamic send/recv selected rows

* update by comment

* fix by comment

Fix #8368

reduce doc build time in travis ci

Improve split and concat op